Callers of flowi4_update_output() never try to update ->flowi4_tos:
* ip_route_connect() updates ->flowi4_tos with its own current
value.
* ip_route_newports() has two users: tcp_v4_connect() and
dccp_v4_connect. Both initialise fl4 with ip_route_connect(), which
in turn sets ->flowi4_tos with RT_TOS(inet_sk(sk)->tos) and
->flowi4_scope based on SOCK_LOCALROUTE.
Then ip_route_newports() updates ->flowi4_tos with
RT_CONN_FLAGS(sk), which is the same as RT_TOS(inet_sk(sk)->tos),
unless SOCK_LOCALROUTE is set on the socket. In that case, the
lowest order bit is set to 1, to eventually inform
ip_route_output_key_hash() to restrict the scope to RT_SCOPE_LINK.
This is equivalent to properly setting ->flowi4_scope as
ip_route_connect() did.
* ip_vs_xmit.c initialises ->flowi4_tos with memset(0), then calls
flowi4_update_output() with tos=0.
* sctp_v4_get_dst() uses the same RT_CONN_FLAGS_TOS() when
initialising ->flowi4_tos and when calling flowi4_update_output().
In the end, ->flowi4_tos never changes. So let's just drop the tos
parameter. This will simplify the conversion of ->flowi4_tos from __u8
to dscp_t.
Signed-off-by: Guillaume Nault <gnault@redhat.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
In the offending fixes commit I mistakenly removed the reply message of
the port new command. I was under impression it is a new port
notification, partly due to the "notify" in the name of the helper
function. Bring the code sending reply with new port message back, this
time putting it directly to devlink_nl_cmd_port_new_doit()
Fixes: c496daeb86 ("devlink: remove duplicate port notification")
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Link: https://lore.kernel.org/r/20230531142025.2605001-1-jiri@resnulli.us
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull networking fixes from Jakub Kicinski:
"Happy Wear a Dress Day.
Fairly standard-sized batch of fixes, accounting for the lack of
sub-tree submissions this week. The mlx5 IRQ fixes are notable, people
were complaining about that. No fires burning.
Current release - regressions:
- eth: mlx5e:
- multiple fixes for dynamic IRQ allocation
- prevent encap offload when neigh update is running
- eth: mana: fix perf regression: remove rx_cqes, tx_cqes counters
Current release - new code bugs:
- eth: mlx5e: DR, add missing mutex init/destroy in pattern manager
Previous releases - always broken:
- tcp: deny tcp_disconnect() when threads are waiting
- sched: prevent ingress Qdiscs from getting installed in random
locations in the hierarchy and moving around
- sched: flower: fix possible OOB write in fl_set_geneve_opt()
- netlink: fix NETLINK_LIST_MEMBERSHIPS length report
- udp6: fix race condition in udp6_sendmsg & connect
- tcp: fix mishandling when the sack compression is deferred
- rtnetlink: validate link attributes set at creation time
- mptcp: fix connect timeout handling
- eth: stmmac: fix call trace when stmmac_xdp_xmit() is invoked
- eth: amd-xgbe: fix the false linkup in xgbe_phy_status
- eth: mlx5e:
- fix corner cases in internal buffer configuration
- drain health before unregistering devlink
- usb: qmi_wwan: set DTR quirk for BroadMobi BM818
Misc:
- tcp: return user_mss for TCP_MAXSEG in CLOSE/LISTEN state if
user_mss set"
* tag 'net-6.4-rc5' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (71 commits)
mptcp: fix active subflow finalization
mptcp: add annotations around sk->sk_shutdown accesses
mptcp: fix data race around msk->first access
mptcp: consolidate passive msk socket initialization
mptcp: add annotations around msk->subflow accesses
mptcp: fix connect timeout handling
rtnetlink: add the missing IFLA_GRO_ tb check in validate_linkmsg
rtnetlink: move IFLA_GSO_ tb check to validate_linkmsg
rtnetlink: call validate_linkmsg in rtnl_create_link
ice: recycle/free all of the fragments from multi-buffer frame
net: phy: mxl-gpy: extend interrupt fix to all impacted variants
net: renesas: rswitch: Fix return value in error path of xmit
net: dsa: mv88e6xxx: Increase wait after reset deactivation
net: ipa: Use correct value for IPA_STATUS_SIZE
tcp: fix mishandling when the sack compression is deferred.
net/sched: flower: fix possible OOB write in fl_set_geneve_opt()
sfc: fix error unwinds in TC offload
net/mlx5: Read embedded cpu after init bit cleared
net/mlx5e: Fix error handling in mlx5e_refresh_tirs
net/mlx5: Ensure af_desc.mask is properly initialized
...
When switching from kthreads to vhost_tasks two bugs were added:
1. The vhost worker tasks's now show up as processes so scripts doing
ps or ps a would not incorrectly detect the vhost task as another
process. 2. kthreads disabled freeze by setting PF_NOFREEZE, but
vhost tasks's didn't disable or add support for them.
To fix both bugs, this switches the vhost task to be thread in the
process that does the VHOST_SET_OWNER ioctl, and has vhost_worker call
get_signal to support SIGKILL/SIGSTOP and freeze signals. Note that
SIGKILL/STOP support is required because CLONE_THREAD requires
CLONE_SIGHAND which requires those 2 signals to be supported.
This is a modified version of the patch written by Mike Christie
<michael.christie@oracle.com> which was a modified version of patch
originally written by Linus.
Much of what depended upon PF_IO_WORKER now depends on PF_USER_WORKER.
Including ignoring signals, setting up the register state, and having
get_signal return instead of calling do_group_exit.
Tidied up the vhost_task abstraction so that the definition of
vhost_task only needs to be visible inside of vhost_task.c. Making
it easier to review the code and tell what needs to be done where.
As part of this the main loop has been moved from vhost_worker into
vhost_task_fn. vhost_worker now returns true if work was done.
The main loop has been updated to call get_signal which handles
SIGSTOP, freezing, and collects the message that tells the thread to
exit as part of process exit. This collection clears
__fatal_signal_pending. This collection is not guaranteed to
clear signal_pending() so clear that explicitly so the schedule()
sleeps.
For now the vhost thread continues to exist and run work until the
last file descriptor is closed and the release function is called as
part of freeing struct file. To avoid hangs in the coredump
rendezvous and when killing threads in a multi-threaded exec. The
coredump code and de_thread have been modified to ignore vhost threads.
Remvoing the special case for exec appears to require teaching
vhost_dev_flush how to directly complete transactions in case
the vhost thread is no longer running.
Removing the special case for coredump rendezvous requires either the
above fix needed for exec or moving the coredump rendezvous into
get_signal.
Fixes: 6e890c5d50 ("vhost: use vhost_tasks for worker threads")
Signed-off-by: Eric W. Biederman <ebiederm@xmission.com>
Co-developed-by: Mike Christie <michael.christie@oracle.com>
Signed-off-by: Mike Christie <michael.christie@oracle.com>
Acked-by: Michael S. Tsirkin <mst@redhat.com>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull firewire fix from Takashi Sakamoto:
"A single patch to use a flexible array rather than a zero-length one"
* tag 'firewire-fixes-6.4-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/ieee1394/linux1394:
firewire: Replace zero-length array with flexible-array member
Zero-length and one-element arrays are deprecated, and we are moving
towards adopting C99 flexible-array members, instead.
Address the following warnings found with GCC-13 and
-fstrict-flex-arrays=3 enabled:
sound/firewire/amdtp-stream.c: In function ‘build_it_pkt_header’:
sound/firewire/amdtp-stream.c:694:17: warning: ‘generate_cip_header’ accessing 8 bytes in a region of size 0 [-Wstringop-overflow=]
694 | generate_cip_header(s, cip_header, data_block_counter, syt);
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
sound/firewire/amdtp-stream.c:694:17: note: referencing argument 2 of type ‘__be32[2]’ {aka ‘unsigned int[2]’}
sound/firewire/amdtp-stream.c:667:13: note: in a call to function ‘generate_cip_header’
667 | static void generate_cip_header(struct amdtp_stream *s, __be32 cip_header[2],
| ^~~~~~~~~~~~~~~~~~~
This helps with the ongoing efforts to tighten the FORTIFY_SOURCE
routines on memcpy() and help us make progress towards globally
enabling -fstrict-flex-arrays=3 [1].
Link: https://github.com/KSPP/linux/issues/21
Link: https://github.com/KSPP/linux/issues/303
Link: https://gcc.gnu.org/pipermail/gcc-patches/2022-October/602902.html [1]
Signed-off-by: Gustavo A. R. Silva <gustavoars@kernel.org>
Reviewed-by: Kees Cook <keescook@chromium.org>
Link: https://lore.kernel.org/r/ZHT0V3SpvHyxCv5W@work
Signed-off-by: Takashi Sakamoto <o-takashi@sakamocchi.jp>
Offloading drivers may report some additional statistics counters, some
of them even suggested by 802.1Q, like TransmissionOverrun.
In my opinion we don't have to limit ourselves to reporting counters
only globally to the Qdisc/interface, especially if the device has more
detailed reporting (per traffic class), since the more detailed info is
valuable for debugging and can help identifying who is exceeding its
time slot.
But on the other hand, some devices may not be able to report both per
TC and global stats.
So we end up reporting both ways, and use the good old ethtool_put_stat()
strategy to determine which statistics are supported by this NIC.
Statistics which aren't set are simply not reported to netlink. For this
reason, we need something dynamic (a nlattr nest) to be reported through
TCA_STATS_APP, and not something daft like the fixed-size and
inextensible struct tc_codel_xstats. A good model for xstats which are a
nlattr nest rather than a fixed struct seems to be cake.
# Global stats
$ tc -s qdisc show dev eth0 root
# Per-tc stats
$ tc -s class show dev eth0
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Acked-by: Vinicius Costa Gomes <vinicius.gomes@intel.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Inspired from struct flow_cls_offload :: cmd, in order for taprio to be
able to report statistics (which is future work), it seems that we need
to drill one step further with the ndo_setup_tc(TC_SETUP_QDISC_TAPRIO)
multiplexing, and pass the command as part of the common portion of the
muxed structure.
Since we already have an "enable" variable in tc_taprio_qopt_offload,
refactor all drivers to check for "cmd" instead of "enable", and reject
every other command except "replace" and "destroy" - to be future proof.
Signed-off-by: Vladimir Oltean <vladimir.oltean@nxp.com>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com> # for lan966x
Acked-by: Kurt Kanzenbach <kurt@linutronix.de> # hellcreek
Reviewed-by: Muhammad Husaini Zulkifli <muhammad.husaini.zulkifli@intel.com>
Reviewed-by: Gerhard Engleder <gerhard@engleder-embedded.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
tcp_gro_complete() function only updates the skb fields related to GRO
and it always returns zero. All the 3 drivers which are using it
do not check for the return value either.
Change it to return void instead which simplifies its callers as
error handing becomes unnecessary.
Signed-off-by: Parav Pandit <parav@nvidia.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Expose netdev trigger modes to make them accessible by LED driver that
will support netdev trigger for hw control.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Some specific LED triggers blink the LED based on events from a device
or subsystem.
For example, an LED could be blinked to indicate a network device is
receiving packets, or a disk is reading blocks. To correctly enable and
request the hw control of the LED, the trigger has to check if the
network interface or block device configured via a /sys/class/led file
match the one the LED driver provide for hw control for.
Provide an API call to get the device which the LED blinks for.
Signed-off-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add an option to permit LED driver to declare support for a specific
trigger to use hw control and setup the LED to blink based on specific
provided modes.
Add APIs for LEDs hw control. These functions will be used to activate
hardware control where a LED will use the provided flags, from an
unique defined supported trigger, to setup the LED to be driven by
hardware.
Add hw_control_is_supported() to ask the LED driver if the requested
mode by the trigger are supported and the LED can be setup to follow
the requested modes.
Deactivate hardware blink control by setting brightness to LED_OFF via
the brightness_set() callback.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add the 'TCA_FLOWER_L2_MISS' netlink attribute that allows user space to
match on packets that encountered a layer 2 miss. The miss indication is
set as metadata in the tc skb extension by the bridge driver upon FDB or
MDB lookup miss and dissected by the flow dissector to the
'FLOW_DISSECTOR_KEY_META' key.
The use of this skb extension is guarded by the 'tc_skb_ext_tc' static
key. As such, enable / disable this key when filters that match on layer
2 miss are added / deleted.
Tested:
# cat tc_skb_ext_tc.py
#!/usr/bin/env -S drgn -s vmlinux
refcount = prog["tc_skb_ext_tc"].key.enabled.counter.value_()
print(f"tc_skb_ext_tc reference count is {refcount}")
# ./tc_skb_ext_tc.py
tc_skb_ext_tc reference count is 0
# tc filter add dev swp1 egress proto all handle 101 pref 1 flower src_mac 00:11:22:33:44:55 action drop
# tc filter add dev swp1 egress proto all handle 102 pref 2 flower src_mac 00:11:22:33:44:55 l2_miss true action drop
# tc filter add dev swp1 egress proto all handle 103 pref 3 flower src_mac 00:11:22:33:44:55 l2_miss false action drop
# ./tc_skb_ext_tc.py
tc_skb_ext_tc reference count is 2
# tc filter replace dev swp1 egress proto all handle 102 pref 2 flower src_mac 00:01:02:03:04:05 l2_miss false action drop
# ./tc_skb_ext_tc.py
tc_skb_ext_tc reference count is 2
# tc filter del dev swp1 egress proto all handle 103 pref 3 flower
# tc filter del dev swp1 egress proto all handle 102 pref 2 flower
# tc filter del dev swp1 egress proto all handle 101 pref 1 flower
# ./tc_skb_ext_tc.py
tc_skb_ext_tc reference count is 0
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Extend the 'FLOW_DISSECTOR_KEY_META' key with a new 'l2_miss' field and
populate it from a field with the same name in the tc skb extension.
This field is set by the bridge driver for packets that incur an FDB or
MDB miss.
The next patch will extend the flower classifier to be able to match on
layer 2 misses.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: Nikolay Aleksandrov <razor@blackwall.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
For EVPN non-DF (Designated Forwarder) filtering we need to be able to
prevent decapsulated traffic from being flooded to a multi-homed host.
Filtering of multicast and broadcast traffic can be achieved using the
following flower filter:
# tc filter add dev bond0 egress pref 1 proto all flower indev vxlan0 dst_mac 01:00:00:00:00:00/01:00:00:00:00:00 action drop
Unlike broadcast and multicast traffic, it is not currently possible to
filter unknown unicast traffic. The classification into unknown unicast
is performed by the bridge driver, but is not visible to other layers
such as tc.
Solve this by adding a new 'l2_miss' bit to the tc skb extension. Clear
the bit whenever a packet enters the bridge (received from a bridge port
or transmitted via the bridge) and set it if the packet did not match an
FDB or MDB entry. If there is no skb extension and the bit needs to be
cleared, then do not allocate one as no extension is equivalent to the
bit being cleared. The bit is not set for broadcast packets as they
never perform a lookup and therefore never incur a miss.
A bit that is set for every flooded packet would also work for the
current use case, but it does not allow us to differentiate between
registered and unregistered multicast traffic, which might be useful in
the future.
To keep the performance impact to a minimum, the marking of packets is
guarded by the 'tc_skb_ext_tc' static key. When 'false', the skb is not
touched and an skb extension is not allocated. Instead, only a
5 bytes nop is executed, as demonstrated below for the call site in
br_handle_frame().
Before the patch:
```
memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
c37b09: 49 c7 44 24 28 00 00 movq $0x0,0x28(%r12)
c37b10: 00 00
p = br_port_get_rcu(skb->dev);
c37b12: 49 8b 44 24 10 mov 0x10(%r12),%rax
memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
c37b17: 49 c7 44 24 30 00 00 movq $0x0,0x30(%r12)
c37b1e: 00 00
c37b20: 49 c7 44 24 38 00 00 movq $0x0,0x38(%r12)
c37b27: 00 00
```
After the patch (when static key is disabled):
```
memset(skb->cb, 0, sizeof(struct br_input_skb_cb));
c37c29: 49 c7 44 24 28 00 00 movq $0x0,0x28(%r12)
c37c30: 00 00
c37c32: 49 8d 44 24 28 lea 0x28(%r12),%rax
c37c37: 48 c7 40 08 00 00 00 movq $0x0,0x8(%rax)
c37c3e: 00
c37c3f: 48 c7 40 10 00 00 00 movq $0x0,0x10(%rax)
c37c46: 00
#ifdef CONFIG_HAVE_JUMP_LABEL_HACK
static __always_inline bool arch_static_branch(struct static_key *key, bool branch)
{
asm_volatile_goto("1:"
c37c47: 0f 1f 44 00 00 nopl 0x0(%rax,%rax,1)
br_tc_skb_miss_set(skb, false);
p = br_port_get_rcu(skb->dev);
c37c4c: 49 8b 44 24 10 mov 0x10(%r12),%rax
```
Subsequent patches will extend the flower classifier to be able to match
on the new 'l2_miss' bit and enable / disable the static key when
filters that match on it are added / deleted.
Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Acked-by: Nikolay Aleksandrov <razor@blackwall.org>
Acked-by: Jakub Kicinski <kuba@kernel.org>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Move port_fn_hw_addr_get/set() from devlink_ops into newly introduced
devlink_port_ops.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Acked-by: Martin Habets <habetsm.xilinx@gmail.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
In devlink, some of the objects have separate ops registered alongside
with the object itself. Port however have ops in devlink_ops structure.
For drivers what register multiple kinds of ports with different ops
this is not convenient. Introduce devlink_port_ops and a set
of functions that allow drivers to pass ops pointer during
port registration.
Signed-off-by: Jiri Pirko <jiri@nvidia.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
The apc->eth_stats.rx_cqes is one per NIC (vport), and it's on the
frequent and parallel code path of all queues. So, r/w into this
single shared variable by many threads on different CPUs creates a
lot caching and memory overhead, hence perf regression. And, it's
not accurate due to the high volume concurrent r/w.
For example, a workload is iperf with 128 threads, and with RPS
enabled. We saw perf regression of 25% with the previous patch
adding the counters. And this patch eliminates the regression.
Since the error path of mana_poll_rx_cq() already has warnings, so
keeping the counter and convert it to a per-queue variable is not
necessary. So, just remove this counter from this high frequency
code path.
Also, remove the tx_cqes counter for the same reason. We have
warnings & other counters for errors on that path, and don't need
to count every normal cqe processing.
Cc: stable@vger.kernel.org
Fixes: bd7fc6e195 ("net: mana: Add new MANA VF performance counters for easier troubleshooting")
Signed-off-by: Haiyang Zhang <haiyangz@microsoft.com>
Reviewed-by: Horatiu Vultur <horatiu.vultur@microchip.com>
Reviewed-by: Jiri Pirko <jiri@nvidia.com>
Link: https://lore.kernel.org/r/1685115537-31675-1-git-send-email-haiyangz@microsoft.com
Signed-off-by: Paolo Abeni <pabeni@redhat.com>
Add lynx_pcs_create_mdiodev() to simplify the creation of the mdio
device associated with lynx PCS. In order to allow lynx_pcs_destroy()
to clean this up, we need to arrange for lynx_pcs_create() to take a
refcount on the mdiodev, and lynx_pcs_destroy() to put it.
Adding the refcounting to lynx_pcs_create()..lynx_pcs_destroy() will
be transparent to existing users of these interfaces.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Reviewed-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Tested-by: Ioana Ciornei <ioana.ciornei@nxp.com>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add xpcs_create_mdiodev() to simplify the creation of the mdio device
associated with the XPCS. In order to allow xpcs_destroy() to clean
this up, we need to arrange for xpcs_create() to take a refcount on
the mdiodev, and xpcs_destroy() to put it.
Adding the refcounting to xpcs_create()..xpcs_destroy() will be
transparent to existing users of these interfaces.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Add two new operations for a mdio device to manage the refcount on the
underlying struct device. This will be used by mdio PCS drivers to
simplify the creation and destruction handling, making it easier for
users to get it correct.
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull tracing fixes from Steven Rostedt:
"User events:
- Use long instead of int for storing the enable set/clear bit, as it
was found that big endian machines could end up using the wrong
bits.
- Split allocating mm and attaching it. This keeps the allocation
separate from the registration and avoids various races.
- Remove RCU locking around pin_user_pages_remote() as that can
schedule. The RCU protection is no longer needed with the above
split of mm allocation and attaching.
- Rename the "link" fields of the various structs to something more
meaningful.
- Add comments around user_event_mm struct usage and locking
requirements.
Timerlat tracer:
- Fix missed wakeup of timerlat thread caused by the timerlat
interrupt triggering when tracing is off. The timer interrupt
handler needs to always wake up the timerlat thread regardless if
tracing is enabled or not, otherwise, it will never wake up.
Histograms:
- Fix regression of breaking the "stacktrace" modifier for variables.
That modifier cannot be used for values, but can be used for
variables that are passed from one histogram to the next. This was
broken when adding the restriction to values as the variable logic
used the same code.
- Rename the special field "stacktrace" to "common_stacktrace".
Special fields (that are not actually part of the event, but can
act just like event fields, like 'comm' and 'timestamp') should be
prefixed with 'common_' for consistency. To keep backward
compatibility, 'stacktrace' can still be used (as with the special
field 'cpu'), but can be overridden if the event has a field called
'stacktrace'.
- Update the synthetic event selftests to use the new name (synthetic
events are created by histograms)
Tracing bootup selftests:
- Reorganize the code to keep artifacts of the selftests not compiled
in when selftests are not configured.
- Add various cond_resched() around the selftest code, as the
softlock watchdog was triggering much more often. It appears that
the kernel runs slower now with full debugging enabled.
- While debugging ftrace with ftrace (using an instance ring buffer
instead of the top level one), I found that the selftests were
disabling prints to the debug instance.
This should not happen, as the selftests only disable printing to
the main buffer as the selftests examine the main buffer to see if
it has what it expects, and prints can make the tests fail.
Make the selftests only disable printing to the toplevel buffer,
and leave the instance buffers alone"
* tag 'trace-v6.4-rc3' of git://git.kernel.org/pub/scm/linux/kernel/git/trace/linux-trace:
tracing: Have function_graph selftest call cond_resched()
tracing: Only make selftest conditionals affect the global_trace
tracing: Make tracing_selftest_running/delete nops when not used
tracing: Have tracer selftests call cond_resched() before running
tracing: Move setting of tracing_selftest_running out of register_tracer()
tracing/selftests: Update synthetic event selftest to use common_stacktrace
tracing: Rename stacktrace field to common_stacktrace
tracing/histograms: Allow variables to have some modifiers
tracing/user_events: Document user_event_mm one-shot list usage
tracing/user_events: Rename link fields for clarity
tracing/user_events: Remove RCU lock while pinning pages
tracing/user_events: Split up mm alloc and attach
tracing/timerlat: Always wakeup the timerlat thread
tracing/user_events: Use long vs int for atomic bit ops
This reverts commit 9828ed3f69.
Sadly, it does seem to cause failures to load modules. Johan Hovold reports:
"This change breaks module loading during boot on the Lenovo Thinkpad
X13s (aarch64).
Specifically it results in indefinite probe deferral of the display
and USB (ethernet) which makes it a pain to debug. Typing in the dark
to acquire some logs reveals that other modules are missing as well"
Since this was applied late as a "let's try this", I'm reverting it
asap, and we can try to figure out what goes wrong later. The excessive
parallel module loading problem is annoying, but not noticeable in
normal situations, and this was only meant as an optimistic workaround
for a user-space bug.
One possible solution may be to do the optimistic exclusive open first,
and then use a lock to serialize loading if that fails.
Reported-by: Johan Hovold <johan@kernel.org>
Link: https://lore.kernel.org/lkml/ZHRpH-JXAxA6DnzR@hovoldconsulting.com/
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull unwinder fixes from Thomas Gleixner:
"A set of unwinder and tooling fixes:
- Ensure that the stack pointer on x86 is aligned again so that the
unwinder does not read past the end of the stack
- Discard .note.gnu.property section which has a pointlessly
different alignment than the other note sections. That confuses
tooling of all sorts including readelf, libbpf and pahole"
* tag 'objtool-urgent-2023-05-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
x86/show_trace_log_lvl: Ensure stack pointer is aligned, again
vmlinux.lds.h: Discard .note.gnu.property section
Pull debugobjects fixes from Thomas Gleixner:
"Two fixes for debugobjects:
- Prevent the allocation path from waking up kswapd.
That's a long standing issue due to the GFP_ATOMIC allocation flag.
As debug objects can be invoked from pretty much any context waking
kswapd can end up in arbitrary lock chains versus the waitqueue
lock
- Correct the explicit lockdep wait-type violation in
debug_object_fill_pool()"
* tag 'core-debugobjects-2023-05-28' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
debugobjects: Don't wake up kswapd from fill_pool()
debugobjects,locking: Annotate debug_object_fill_pool() wait type violation
Pull xen fixes from Juergen Gross:
- a double free fix in the Xen pvcalls backend driver
- a fix for a regression causing the MSI related sysfs entries to not
being created in Xen PV guests
- a fix in the Xen blkfront driver for handling insane input data
better
* tag 'for-linus-6.4-rc4-tag' of git://git.kernel.org/pub/scm/linux/kernel/git/xen/tip:
x86/pci/xen: populate MSI sysfs entries
xen/pvcalls-back: fix double frees with pvcalls_new_active_socket()
xen/blkfront: Only check REQ_FUA for writes
Daniel Borkmann says:
====================
pull-request: bpf-next 2023-05-26
We've added 54 non-merge commits during the last 10 day(s) which contain
a total of 76 files changed, 2729 insertions(+), 1003 deletions(-).
The main changes are:
1) Add the capability to destroy sockets in BPF through a new kfunc,
from Aditi Ghag.
2) Support O_PATH fds in BPF_OBJ_PIN and BPF_OBJ_GET commands,
from Andrii Nakryiko.
3) Add capability for libbpf to resize datasec maps when backed via mmap,
from JP Kobryn.
4) Move all the test kfuncs for CI out of the kernel and into bpf_testmod,
from Jiri Olsa.
5) Big batch of xsk selftest improvements to prep for multi-buffer testing,
from Magnus Karlsson.
6) Show the target_{obj,btf}_id in tracing link's fdinfo and dump it
via bpftool, from Yafang Shao.
7) Various misc BPF selftest improvements to work with upcoming LLVM 17,
from Yonghong Song.
8) Extend bpftool to specify netdevice for resolving XDP hints,
from Larysa Zaremba.
9) Document masking in shift operations for the insn set document,
from Dave Thaler.
10) Extend BPF selftests to check xdp_feature support for bond driver,
from Lorenzo Bianconi.
* tag 'for-netdev' of https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next: (54 commits)
bpf: Fix bad unlock balance on freeze_mutex
libbpf: Ensure FD >= 3 during bpf_map__reuse_fd()
libbpf: Ensure libbpf always opens files with O_CLOEXEC
selftests/bpf: Check whether to run selftest
libbpf: Change var type in datasec resize func
bpf: drop unnecessary bpf_capable() check in BPF_MAP_FREEZE command
libbpf: Selftests for resizing datasec maps
libbpf: Add capability for resizing datasec maps
selftests/bpf: Add path_fd-based BPF_OBJ_PIN and BPF_OBJ_GET tests
libbpf: Add opts-based bpf_obj_pin() API and add support for path_fd
bpf: Support O_PATH FDs in BPF_OBJ_PIN and BPF_OBJ_GET commands
libbpf: Start v1.3 development cycle
bpf: Validate BPF object in BPF_OBJ_PIN before calling LSM
bpftool: Specify XDP Hints ifname when loading program
selftests/bpf: Add xdp_feature selftest for bond device
selftests/bpf: Test bpf_sock_destroy
selftests/bpf: Add helper to get port using getsockname
bpf: Add bpf_sock_destroy kfunc
bpf: Add kfunc filter function to 'struct btf_kfunc_id_set'
bpf: udp: Implement batching for sockets iterator
...
====================
Link: https://lore.kernel.org/r/20230526222747.17775-1-daniel@iogearbox.net
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
Pull ARM SoC fixes from Arnd Bergmann:
"There have not been a lot of fixes for for the soc tree in 6.4, but
these have been sitting here for too long.
For the devicetree side, there is one minor warning fix for vexpress,
the rest all all for the the NXP i.MX platforms: SoC specific bugfixes
for the iMX8 clocks and its USB-3.0 gadget device, as well as board
specific fixes for regulators and the phy on some of the i.MX boards.
The microchip risc-v and arm32 maintainers now also add a shared
maintainer file entry for the arm64 parts.
The remaining fixes are all for firmware drivers, addressing mistakes
in the optee, scmi and ff-a firmware driver implementation, mostly in
the error handling code, incorrect use of the alloc_workqueue()
interface in SCMI, and compatibility with corner cases of the firmware
implementation"
* tag 'arm-fixes-6.4-1' of git://git.kernel.org/pub/scm/linux/kernel/git/soc/soc:
MAINTAINERS: update arm64 Microchip entries
arm64: dts: imx8: fix USB 3.0 Gadget Failure in QM & QXPB0 at super speed
dt-binding: cdns,usb3: Fix cdns,on-chip-buff-size type
arm64: dts: colibri-imx8x: delete adc1 and dsp
arm64: dts: colibri-imx8x: fix iris pinctrl configuration
arm64: dts: colibri-imx8x: move pinctrl property from SoM to eval board
arm64: dts: colibri-imx8x: fix eval board pin configuration
arm64: dts: imx8mp: Fix video clock parents
ARM: dts: imx6qdl-mba6: Add missing pvcie-supply regulator
ARM: dts: imx6ull-dhcor: Set and limit the mode for PMIC buck 1, 2 and 3
arm64: dts: imx8mn-var-som: fix PHY detection bug by adding deassert delay
arm64: dts: imx8mn: Fix video clock parents
firmware: arm_ffa: Set reserved/MBZ fields to zero in the memory descriptors
firmware: arm_ffa: Fix FFA device names for logical partitions
firmware: arm_ffa: Fix usage of partition info get count flag
firmware: arm_ffa: Check if ffa_driver remove is present before executing
arm64: dts: arm: add missing cache properties
ARM: dts: vexpress: add missing cache properties
firmware: arm_scmi: Fix incorrect alloc_workqueue() invocation
optee: fix uninited async notif value
Pull drm fixes from Dave Airlie:
"This week's collection is pretty spread out, accel/qaic has a bunch of
fixes, amdgpu, then lots of single fixes across a bunch of places.
core:
- fix drmm_mutex_init lock class
mgag200:
- fix gamma lut initialisation
pl111:
- fix FB depth on IMPD-1 framebuffer
amdgpu:
- Fix missing BO unlocking in KIQ error path
- Avoid spurious secure display error messages
- SMU13 fix
- Fix an OD regression
- GPU reset display IRQ warning fix
- MST fix
radeon:
- Fix a DP regression
i915:
- PIPEDMC disabling fix for bigjoiner config
panel:
- fix aya neo air plus quirk
sched:
- remove redundant NULL check
qaic:
- fix NNC message corruption
- Grab ch_lock during QAIC_ATTACH_SLICE_BO
- Flush the transfer list again
- Validate if BO is sliced before slicing
- Validate user data before grabbing any lock
- initialize ret variable to 0
- silence some uninitialized variable warnings"
* tag 'drm-fixes-2023-05-26' of git://anongit.freedesktop.org/drm/drm:
drm/amd/display: Have Payload Properly Created After Resume
drm/amd/display: Fix warning in disabling vblank irq
drm/amd/pm: Fix output of pp_od_clk_voltage
drm/amd/pm: add missing NotifyPowerSource message mapping for SMU13.0.7
drm/radeon: reintroduce radeon_dp_work_func content
drm/amdgpu: don't enable secure display on incompatible platforms
drm:amd:amdgpu: Fix missing buffer object unlock in failure path
accel/qaic: Fix NNC message corruption
accel/qaic: Grab ch_lock during QAIC_ATTACH_SLICE_BO
accel/qaic: Flush the transfer list again
accel/qaic: Validate if BO is sliced before slicing
accel/qaic: Validate user data before grabbing any lock
accel/qaic: initialize ret variable to 0
drm/i915: Fix PIPEDMC disabling for a bigjoiner configuration
drm: fix drmm_mutex_init()
drm/sched: Remove redundant check
drm: panel-orientation-quirks: Change Air's quirk to support Air Plus
accel/qaic: silence some uninitialized variable warnings
drm/pl111: Fix FB depth on IMPD-1 framebuffer
drm/mgag200: Fix gamma lut not initialized.
Arm FF-A fixes for v6.4
Quite a few fixes to address set of assorted issues:
1. NULL pointer dereference if the ffa driver doesn't provide remove()
callback as it is currently executed unconditionally
2. FF-A core probe failure on systems with v1.0 firmware as the new
partition info get count flag is used unconditionally
3. Failure to register more than one logical partition or service within
the same physical partition as the device name contains only VM ID
which will be same for all but each will have unique UUID.
4. Rejection of certain memory interface transmissions by the receivers
(secure partitions) as few MBZ fields are non-zero due to lack of
explicit re-initialization of those fields
* tag 'ffa-fixes-6.4' of git://git.kernel.org/pub/scm/linux/kernel/git/sudeep.holla/linux:
firmware: arm_ffa: Set reserved/MBZ fields to zero in the memory descriptors
firmware: arm_ffa: Fix FFA device names for logical partitions
firmware: arm_ffa: Fix usage of partition info get count flag
firmware: arm_ffa: Check if ffa_driver remove is present before executing
Link: https://lore.kernel.org/r/20230509143453.1188753-1-sudeep.holla@arm.com
Signed-off-by: Arnd Bergmann <arnd@arndb.de>
Currently TCP_SYNQ_INTERVAL is defined but never used.
According to "git log -S TCP_SYNQ_INTERVAL net-next/main" it seems
the last references to TCP_SYNQ_INTERVAL were removed by 2015
commit fa76ce7328 ("inet: get rid of central tcp/dccp listener timer")
Signed-off-by: Neal Cardwell <ncardwell@google.com>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Reviewed-by: Eric Dumazet <edumazet@google.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Add DSA support for the phylink mac_prepare() and mac_finish() calls.
These were introduced as part of the PCS support to allow MACs to
perform preparatory steps prior to configuration, and finalisation
steps after the MAC and PCS has been configured.
Introducing phylink_pcs support to the mv88e6xxx DSA driver needs some
code moved out of its mac_config() stage into the mac_prepare() and
mac_finish() stages, and this commit facilitates such code in DSA
drivers.
Reviewed-by: Andrew Lunn <andrew@lunn.ch>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>
Reviewed-by: Simon Horman <simon.horman@corigine.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Saeed Mahameed says:
====================
mlx5 fixes 2023-05-24
This series includes bug fixes for the mlx5 driver.
* tag 'mlx5-fixes-2023-05-24' of git://git.kernel.org/pub/scm/linux/kernel/git/saeed/linux:
Documentation: net/mlx5: Wrap notes in admonition blocks
Documentation: net/mlx5: Add blank line separator before numbered lists
Documentation: net/mlx5: Use bullet and definition lists for vnic counters description
Documentation: net/mlx5: Wrap vnic reporter devlink commands in code blocks
net/mlx5: Fix check for allocation failure in comp_irqs_request_pci()
net/mlx5: DR, Add missing mutex init/destroy in pattern manager
net/mlx5e: Move Ethernet driver debugfs to profile init callback
net/mlx5e: Don't attach netdev profile while handling internal error
net/mlx5: Fix post parse infra to only parse every action once
net/mlx5e: Use query_special_contexts cmd only once per mdev
net/mlx5: fw_tracer, Fix event handling
net/mlx5: SF, Drain health before removing device
net/mlx5: Drain health before unregistering devlink
net/mlx5e: Do not update SBCM when prio2buffer command is invalid
net/mlx5e: Consider internal buffers size in port buffer calculations
net/mlx5e: Prevent encap offload when neigh update is running
net/mlx5e: Extract remaining tunnel encap code to dedicated file
====================
Link: https://lore.kernel.org/r/20230525034847.99268-1-saeed@kernel.org
Signed-off-by: Jakub Kicinski <kuba@kernel.org>
It turns out that udev under certain circumstances will concurrently try
to load the same modules over-and-over excessively. This isn't a kernel
bug, but it ends up affecting the kernel, to the point that under
certain circumstances we can fail to boot, because the kernel uses a lot
of memory to read all the module data all at once.
Note that it isn't a memory leak, it's just basically a thundering herd
problem happening at bootup with a lot of CPUs, with the worst cases
then being pretty bad.
Admittedly the worst situations are somewhat contrived: lots and lots of
CPUs, not a lot of memory, and KASAN enabled to make it all slower and
as such (unintentionally) exacerbate the problem.
Luis explains: [1]
"My best assessment of the situation is that each CPU in udev ends up
triggering a load of duplicate set of modules, not just one, but *a
lot*. Not sure what heuristics udev uses to load a set of modules per
CPU."
Petr Pavlu chimes in: [2]
"My understanding is that udev workers are forked. An initial kmod
context is created by the main udevd process but no sharing happens
after the fork. It means that the mentioned memory pool logic doesn't
really kick in.
Multiple parallel load requests come from multiple udev workers, for
instance, each handling an udev event for one CPU device and making
the exactly same requests as all others are doing at the same time.
The optimization idea would be to recognize these duplicate requests
at the udevd/kmod level and converge them"
Note that module loading has tried to mitigate this issue before, see
for example commit 064f4536d1 ("module: avoid allocation if module is
already present and ready"), which has a few ASCII graphs on memory use
due to this same issue.
However, while that noticed that the module was already loaded, and
exited with an error early before spending any more time on setting up
the module, it didn't handle the case of multiple concurrent module
loads all being active - but not complete - at the same time.
Yes, one of them will eventually win the race and finalize its copy, and
the others will then notice that the module already exists and error
out, but while this all happens, we have tons of unnecessary concurrent
work being done.
Again, the real fix is for udev to not do that (maybe it should use
threads instead of fork, and have actual shared data structures and not
cause duplicate work). That real fix is apparently not trivial.
But it turns out that the kernel already has a pretty good model for
dealing with concurrent access to the same file: the i_writecount of the
inode.
In fact, the module loading already indirectly uses 'i_writecount' ,
because 'kernel_file_read()' will in fact do
ret = deny_write_access(file);
if (ret)
return ret;
...
allow_write_access(file);
around the read of the file data. We do not allow concurrent writes to
the file, and return -ETXTBUSY if the file was open for writing at the
same time as the module data is loaded from it.
And the solution to the reader concurrency problem is to simply extend
this "no concurrent writers" logic to simply be "exclusive access".
Note that "exclusive" in this context isn't really some absolute thing:
it's only exclusion from writers and from other "special readers" that
do this writer denial. So we simply introduce a variation of that
"deny_write_access()" logic that not only denies write access, but also
requires that this is the _only_ such access that denies write access.
Which means that you can't start loading a module that is already being
loaded as a module by somebody else, or you will get the same -ETXTBSY
error that you would get if there were writers around.
[ It also means that you can't try to load a currently executing
executable as a module, for the same reason: executables do that same
"deny_write_access()" thing, and that's obviously where the whole
ETXTBSY logic traditionally came from.
This is not a problem for kernel modules, since the set of normal
executable files and kernel module files is entirely disjoint. ]
This new function is called "exclusive_deny_write_access()", and the
implementation is trivial, in that it's just an atomic decrement of
i_writecount if it was 0 before.
To use that new exclusivity check, all we then do is wrap the module
loading with that exclusive_deny_write_access()() / allow_write_access()
pair. The actual patch is a bit bigger than that, because we want to
surround not just the "load file data" part, but the whole module setup,
to get maximum exclusion.
So this ends up splitting up "finit_module()" into a few helper
functions to make it all very clear and legible.
In Luis' test-case (bringing up 255 vcpu's in a virtual machine [3]),
the "wasted vmalloc" space (ie module data read into a vmalloc'ed area
in order to be loaded as a module, but then discarded because somebody
else loaded the same module instead) dropped from 1.8GiB to 474kB. Yes,
that's gigabytes to kilobytes.
It doesn't drop completely to zero, because even with this change, you
can still end up having completely serial pointless module loads, where
one udev process has loaded a module fully (and thus the kernel has
released that exclusive lock on the module file), and then another udev
process tries to load the same module again.
So while we cannot fully get rid of the fundamental bug in user space,
we _can_ get rid of the excessive concurrent thundering herd effect.
A couple of final side notes on this all:
- This tweak only affects the "finit_module()" system call, which gives
the kernel a file descriptor with the module data.
You can also just feed the module data as raw data from user space
with "init_module()" (note the lack of 'f' at the beginning), and
obviously for that case we do _not_ have any "exclusive read" logic.
So if you absolutely want to do things wrong in user space, and try
to load the same module multiple times, and error out only later when
the kernel ends up saying "you can't load the same module name
twice", you can still do that.
And in fact, some distros will do exactly that, because they will
uncompress the kernel module data in user space before feeding it to
the kernel (mainly because they haven't started using the new kernel
side decompression yet).
So this is not some absolute "you can't do concurrent loads of the
same module". It's literally just a very simple heuristic that will
catch it early in case you try to load the exact same module file at
the same time, and in that case avoid a potentially nasty situation.
- There is another user of "deny_write_access()": the verity code that
enables fs-verity on a file (the FS_IOC_ENABLE_VERITY ioctl).
If you use fs-verity and you care about verifying the kernel modules
(which does make sense), you should do it *before* loading said
kernel module. That may sound obvious, but now the implementation
basically requires it. Because if you try to do it concurrently, the
kernel may refuse to load the module file that is being set up by the
fs-verity code.
- This all will obviously mean that if you insist on loading the same
module in parallel, only one module load will succeed, and the others
will return with an error.
That was true before too, but what is different is that the -ETXTBSY
error can be returned *before* the success case of another process
fully loading and instantiating the module.
Again, that might sound obvious, and it is indeed the whole point of
the whole change: we are much quicker to notice the whole "you're
already in the process of loading this module".
So it's very much intentional, but it does mean that if you just
spray the kernel with "finit_module()", and expect that the module is
immediately loaded afterwards without checking the return value, you
are doing something horribly horribly wrong.
I'd like to say that that would never happen, but the whole _reason_
for this commit is that udev is currently doing something horribly
horribly wrong, so ...
Link: https://lore.kernel.org/all/ZEGopJ8VAYnE7LQ2@bombadil.infradead.org/ [1]
Link: https://lore.kernel.org/all/23bd0ce6-ef78-1cd8-1f21-0e706a00424a@suse.com/ [2]
Link: https://lore.kernel.org/lkml/ZG%2Fa+nrt4%2FAAUi5z@bombadil.infradead.org/ [3]
Cc: Greg Kroah-Hartman <gregkh@linuxfoundation.org>
Cc: Lucas De Marchi <lucas.demarchi@intel.com>
Cc: Petr Pavlu <petr.pavlu@suse.com>
Tested-by: Luis Chamberlain <mcgrof@kernel.org>
Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Pull vfs fixes from Christian Brauner:
- During the acl rework we merged this cycle the generic_listxattr()
helper had to be modified in a way that in principle it would allow
for POSIX ACLs to be reported. At least that was the impression we
had initially. Because before the acl rework POSIX ACLs would be
reported if the filesystem did have POSIX ACL xattr handlers in
sb->s_xattr. That logic changed and now we can simply check whether
the superblock has SB_POSIXACL set and if the inode has
inode->i_{default_}acl set report the appropriate POSIX ACL name.
However, we didn't realize that generic_listxattr() was only ever
used by two filesystems. Both of them don't support POSIX ACLs via
sb->s_xattr handlers and so never reported POSIX ACLs via
generic_listxattr() even if they raised SB_POSIXACL and did contain
inodes which had acls set. The example here is nfs4.
As a result, generic_listxattr() suddenly started reporting POSIX
ACLs when it wouldn't have before. Since SB_POSIXACL implies that the
umask isn't stripped in the VFS nfs4 can't just drop SB_POSIXACL from
the superblock as it would also alter umask handling for them.
So just have generic_listxattr() not report POSIX ACLs as it never
did anyway. It's documented as such.
- Our SB_* flags currently use a signed integer and we shift the last
bit causing UBSAN to complain about undefined behavior. Switch to
using unsigned. While the original patch used an explicit unsigned
bitshift it's now pretty common to rely on the BIT() macro in a lot
of headers nowadays. So the patch has been adjusted to use that.
- Add Namjae as ntfs reviewer. They're already active this cycle so
let's make it explicit right now.
* tag 'vfs/v6.4-rc3/misc.fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/vfs/vfs:
ntfs: Add myself as a reviewer
fs: don't call posix_acl_listxattr in generic_listxattr
fs: fix undefined behavior in bit shift for SB_NOUSER
Pull networking fixes from Paolo Abeni:
"Including fixes from bluetooth and bpf.
Current release - regressions:
- net: fix skb leak in __skb_tstamp_tx()
- eth: mtk_eth_soc: fix QoS on DSA MAC on non MTK_NETSYS_V2 SoCs
Current release - new code bugs:
- handshake:
- fix sock->file allocation
- fix handshake_dup() ref counting
- bluetooth:
- fix potential double free caused by hci_conn_unlink
- fix UAF in hci_conn_hash_flush
Previous releases - regressions:
- core: fix stack overflow when LRO is disabled for virtual
interfaces
- tls: fix strparser rx issues
- bpf:
- fix many sockmap/TCP related issues
- fix a memory leak in the LRU and LRU_PERCPU hash maps
- init the offload table earlier
- eth: mlx5e:
- do as little as possible in napi poll when budget is 0
- fix using eswitch mapping in nic mode
- fix deadlock in tc route query code
Previous releases - always broken:
- udplite: fix NULL pointer dereference in __sk_mem_raise_allocated()
- raw: fix output xfrm lookup wrt protocol
- smc: reset connection when trying to use SMCRv2 fails
- phy: mscc: enable VSC8501/2 RGMII RX clock
- eth: octeontx2-pf: fix TSOv6 offload
- eth: cdc_ncm: deal with too low values of dwNtbOutMaxSize"
* tag 'net-6.4-rc4' of git://git.kernel.org/pub/scm/linux/kernel/git/netdev/net: (79 commits)
udplite: Fix NULL pointer dereference in __sk_mem_raise_allocated().
net: phy: mscc: enable VSC8501/2 RGMII RX clock
net: phy: mscc: remove unnecessary phydev locking
net: phy: mscc: add support for VSC8501
net: phy: mscc: add VSC8502 to MODULE_DEVICE_TABLE
net/handshake: Enable the SNI extension to work properly
net/handshake: Unpin sock->file if a handshake is cancelled
net/handshake: handshake_genl_notify() shouldn't ignore @flags
net/handshake: Fix uninitialized local variable
net/handshake: Fix handshake_dup() ref counting
net/handshake: Remove unneeded check from handshake_dup()
ipv6: Fix out-of-bounds access in ipv6_find_tlv()
net: ethernet: mtk_eth_soc: fix QoS on DSA MAC on non MTK_NETSYS_V2 SoCs
docs: netdev: document the existence of the mail bot
net: fix skb leak in __skb_tstamp_tx()
r8169: Use a raw_spinlock_t for the register locks.
page_pool: fix inconsistency for page_pool_ring_[un]lock()
bpf, sockmap: Test progs verifier error with latest clang
bpf, sockmap: Test FIONREAD returns correct bytes in rx buffer with drops
bpf, sockmap: Test FIONREAD returns correct bytes in rx buffer
...
